[cmccomb]

Scope

Record one more independent physical-default OSWEC irregular selected-heading refinement. The new public WEC.run trajectory uses the pinned 500-component PM sea, 241-heading bank, and 0.000625 s step through 250 s. It takes no saved MATLAB pose, force, or heading event as input.

Measured result

The 0.00125-to-0.000625 s Python comparison passes the existing yaw, yaw-speed, PTO-torque, and relative-work limits through 120 s, then fails yaw and work by 140 s. The first yaw-limit breach is 139.7725 s. At 250 s its maximum yaw difference is 0.01966 rad and relative PTO-work difference is 0.965%; the preceding 0.0025-to-0.00125 s comparison measured 0.01256 rad and 0.0552%. The new path also fails the unchanged direct published-MATLAB limits by 140 s. This is nonmonotonic selected-heading behavior, not evidence of physical convergence.

The evidence JSON records the fixed limits, all four windows, input and output hashes, and the plot includes this third Python step pair. The case remains mixed_modes; RM3/OSWEC/Sphere and total strict badge counts do not change. A separately running same-bank MATLAB quarter-step audit will be evaluated when it completes.

Checks

• python3 tools/render_parity_badge.py --check
• git diff --check
• 63-row case inventory and evidence JSON consistency checks
• Local 250 s Python process completed successfully; source and comparison inputs matched their recorded hashes